What This Is
Why This Programme Is Different
Most courses teach you to use AI tools. This programme teaches you to build them. Every concept is applied immediately in a live coding environment. You will write code in every session — there are no slide-only classes.
The RPT 60/30/10 Model
60%
Concept Mastery
Live demo-driven lectures. Every concept is illustrated with running code on the instructor's screen.
30%
Hands-On Lab
You build the same system just discussed — in your own Colab notebook, with TA support available.
10%
Peer Review
Structured critique of each other's outputs. Accountability is built into every session.
Engineered for Impact
Curriculum designed backwards from real AI engineering job requirements.
Mentored by Veterans
Every session led by practitioners who have shipped AI to production.
Build Community
Teams leave with shared codebases and a culture of AI experimentation.
Outcome First
Success = what your team ships after the programme, not attendance.
Programme Snapshot
| Level | Intermediate (Python required) |
| Format | Instructor-led · Lab-first · Offline |
| Effort | 24 hrs/week × 6 weeks |
| Audience | Btech/MCA students (ITER) |
| Certificate | RPT Verified — at ≥ 70% overall |
| Contact | contact@rptcohort.com |
8 Modules · 45 Classes · 90 Hours
Your 6-Week Learning Journey
Each module builds on the previous. Skip nothing — every lab produces an artefact that feeds into your capstone project.
M1
Foundations of Generative AI
Transformer architecture (attention, multi-head, positional encoding) · Tokenization (BPE, WordPiece) · Contextual embeddings vs Word2Vec · Prompt engineering from zero-shot to tree-of-thought · LLM parameter tuning · Cost-quality benchmarking (MMLU, HumanEval, LMSYS Arena)
M2
LLM APIs & Developer Tooling
OpenAI & Anthropic SDKs · Streaming & async patterns · LangChain (LCEL, memory types) · Structured outputs & function calling · LangChain agents with Tavily, Python REPL, Wikipedia tools · RouterChain
Labs: Streaming CLI chatbot with rate-limit retry · LangChain PDF Q&A app · Structured extraction pipeline (20 job postings → Pydantic → CSV) · Research agent
M3
Vector Databases & Semantic Search
Embedding families (OpenAI, BGE-M3, SBERT) · FAISS index types (Flat/IVF/HNSW/PQ) · ChromaDB & Weaviate · Pinecone (serverless) & Qdrant · Hybrid search · Cross-encoder reranking (Cohere, BGE) · Recall@k benchmarking · NDCG, MRR, Hit Rate metrics
Labs: Multi-vector-DB benchmark · Hybrid search implementation · Reranking pipeline comparison
M4
Retrieval-Augmented Generation (RAG)
RAG vs fine-tuning vs long-context decision matrix · Full RAG pipeline (ingest → chunk → embed → store → retrieve → augment → generate → cite) · Chunking strategies · HyDE, multi-query, step-back prompting · RAGAS evaluation (faithfulness, relevancy, context precision) · GPTCache semantic caching · LangSmith/Langfuse observability · Multi-tenant isolation
Labs: End-to-end RAG chatbot · Chunking strategy ablation · RAGAS evaluation dashboard · Production-ready RAG with caching & tracing
M5
Fine-Tuning & Model Customisation
Fine-tuning decision matrix · Dataset formats (Alpaca, ShareGPT, ChatML) · LoRA mathematics (W = W₀ + BA) · QLoRA 4-bit NF4 quantization · PEFT + SFTTrainer · Flash Attention 2 · W&B monitoring · ROUGE/BERTScore/G-Eval evaluation · GGUF quantization · HuggingFace Hub deployment · Multimodal: CLIP, LLaVA, GPT-4o vision · Whisper audio · SDXL / DALL·E 3
Labs: Fine-tune Mistral-7B with LoRA · QLoRA run on Llama-3-8B · Deploy model card + Gradio demo to HF Hub
M6
Building Production AI Applications
ReAct agents from scratch · LangGraph (StateGraph, TypedDict, cyclic graphs, MemorySaver) · Multi-agent systems (CrewAI + AutoGen) · Gradio Blocks & Chatbot · Streamlit session_state · FastAPI SSE streaming · HF Spaces deployment · Guardrails (Presidio PII, prompt injection detection, NeMo) · LLM evaluation & CI (GitHub Actions) · Ollama / vLLM / llama.cpp · Production system design (LLM gateway, async queue, caching tier, cost at 10K users)
Labs: ReAct agent with annotated traces · LangGraph stateful workflow · Multi-agent research system · Full-stack Gradio app deployed to HF Spaces · CI eval pipeline
M7
Ethics, Safety & Responsible AI
RLHF, Constitutional AI, RLAIF · Jailbreak taxonomy (direct, indirect, many-shot, multi-turn) · Prompt injection in agentic systems · Bias types & detection · Hallucination taxonomy (SelfCheckGPT, FActScore) · GDPR for AI · EU AI Act risk tiers (prohibited / high-risk / limited / minimal) · NIST AI RMF · ISO/IEC 42001 · Model cards · ML CO₂ Impact calculator
Labs: Structured red-team exercise · Bias audit report · EU AI Act risk classification of 3 real AI systems · Write your own model card
M8
Industry Applications & MCP
Healthcare Gen AI (clinical NLP, LLaVA-Med, AlphaFold + LLMs) · Software engineering (Copilot productivity, code review bots) · Finance (Bloomberg GPT, earnings analysis, risk reports) · Model Context Protocol (MCP): host/client/server architecture, FastMCP, remote deployment · Career prep: Gen AI role landscape, portfolio review, common technical interview questions
Labs: Custom MCP server implementation · Industry case study analysis · Mock technical interview round
44 Hours · 5 Phases · Teams of 2–3
The Capstone Project
Running in parallel with instruction from Day 1, your capstone is not a tutorial clone — it is an original AI application you design, build, evaluate, and publicly deploy. This becomes your portfolio.
①
Ideation & Planning
4 sessions · 8 hrs
Team formation, Miro brainstorming, feasibility check, Excalidraw architecture diagram, peer review, instructor sign-off. Deliverable: approved proposal v2 + GitHub repo.
②
Data & Backend Build
6 sessions · 12 hrs
500+ row dataset, core LLM integration with SQLite logging, RAG/agent/fine-tune core feature, RAGAS baseline, guardrails, LangSmith tracing, ≥15 pytest tests.
③
Frontend & Deployment
4 sessions · 8 hrs
Gradio/Streamlit UI with streaming & file upload, HuggingFace Spaces deployment, latency profiling, semantic caching (≥25% reduction), security audit, model card.
④
Evaluation & Report
4 sessions · 8 hrs
30-item automated eval, human eval (3 raters), ablation study, IEEE-format project report (8–12 pages, Overleaf) with architecture diagrams and evaluation tables.
⑤
Demo & Submission
4 sessions · 8 hrs
8-min live demo, 10-slide deck, 3-min video (OBS/Loom), full submission package: GitHub + HF Spaces + report PDF + slides + eval CSV + model card + peer reviews.
Final Submission Package — What You Submit
① GitHub repo (clean README, demo GIF, requirements.txt, docstrings)
② HuggingFace Spaces live URL (public, 3 demo examples)
③ IEEE-format project report PDF (8–12 pages)
④ 10-slide deck (PDF) + 3-min demo video (MP4/YouTube)
⑤ Evaluation results CSV + model card (fine-tune projects)
3 Industry Practitioners
Your Instruction Team
Every facilitator has built and deployed Gen AI systems in production. They have debugged the errors your team will face — before you face them.
RT
Rudra Prasad Tripathy
Lead Instructor · Enterprise Architect
20+ years · Capgemini, TCS, HCLTech, Emirates
Seasoned technologist with cross-domain depth in Gen AI, Cloud, Data Engineering, Security, and Supply Chain. Designed curriculum for MCA and engineering programmes across industry-professional workshops.
Industry Track Record
Customer segmentation for the world's largest airlines
AI+ productivity transformation for two of Europe's top automotive brands
Intelligent bot deployment for a leading logistics company in Eastern USA
SS
Santa Swarup Samantaray
Solution Architect
21+ years · Emids, Cisco, Accenture, Wipro
Expert in Scala, Python, GenAI, Kafka, Spark/PySpark, and cloud platforms (AWS & Azure). Active interview panellist at Cisco, Accenture, and Emids Technologies across 10+ projects. International experience: Japan (2008–09), UK (2017).
Key Expertise
Big Data: Spark, PySpark, Kafka, Cassandra, MongoDB
AWS: EC2, S3, EMR, Glue, Lambda, Redshift, SageMaker
Azure: Databricks, Data Factory, CosmosDB, Unity Catalog
UK
Udit Kumar Nayak
Technical Architect · AI & Data Analytics
12+ years · MCA, Sambalpur University
Technical Architect with expertise in LLM tuning, RAG, FastAPI, NLP, and AI Automation. Proficient in Power BI, SQL, ETL processes, KPI development, and integrating AI services into web and analytics solutions.
Key Projects
MAMTA-HIMC Maternal & Child Health Analytics Dashboard
The World Bank — Gender Assessment & Women Safety Analytics
George Washington University — mRANI Analytics Project
20+ Industry Tools
The Production Toolchain
Every tool in this programme is in active use at the world's leading AI engineering teams. You will leave fluent in the stack — not just aware of it.
LLM APIs
OpenAI (GPT-4o, o1)
Anthropic (Claude Sonnet/Opus)
Google Gemini 1.5 Pro
HuggingFace Inference
Orchestration
LangChain + LCEL
LangGraph
CrewAI
AutoGen
Vector DBs
FAISS (Flat/IVF/HNSW/PQ)
ChromaDB + Weaviate
Pinecone (serverless)
Qdrant
Evaluation
RAGAS + G-Eval
LangSmith + Langfuse
pytest + GitHub Actions
Weights & Biases
Deployment
Gradio + Streamlit
FastAPI (SSE streaming)
HuggingFace Spaces
Google Colab Pro A100
Local / Edge
Ollama (Modelfile)
vLLM (PagedAttention)
llama.cpp (GGUF)
MCP Protocol + FastMCP
Safety
Presidio (PII detection)
NeMo Guardrails
OpenAI Moderation API
Guardrails AI
Collaboration
GitHub + Notion
Miro + Excalidraw
Padlet + Overleaf
PEFT + SFTTrainer + TRL
Three Configurations
How We Deliver
The same hands-on, mentor-driven standard across every format. Zero slide-only sessions — ever.
Offline
On-site at ITER. Instructor + TAs physically present. Pair programming, whiteboarding, and real-time code reviews. Best for cohort bonding and maximum learning velocity.
Hybrid
Blended on-site + virtual. Core modules in person; applied labs and reviews online. Ideal for distributed teams or rolling cohorts across multiple locations.
Online Live
Fully virtual, fully live. Shared Colab labs, Padlet walls, and LangSmith dashboards. No passive watching — every session is interactive.
Grading Structure
How You Are Assessed
Assessment is continuous and multi-modal, balancing individual accountability with team collaboration. All components are designed around real engineering tasks with tangible, shareable deliverables.
| Component | Weight | Format | Graded |
|---|---|---|---|
| Class Performance Labs + weekly assignments |
40%
|
8 graded lab notebooks + 8 weekly deliverables (pipelines, GitHub repos, reports) | Individual |
| Capstone Project Team submission |
40%
|
GitHub repo + HF Spaces app + IEEE report + eval CSV + model card. 5 dimensions: functionality, innovation, evaluation rigor, code quality, documentation. | Team + Individual |
| Presentations & Viva 5 events across 6 weeks |
20%
|
Mid-course concept presentation (6%) · Proposal pitch (3%) · Progress demo (3%) · Capstone Demo Day (5%) · Individual viva (3%) | Individual |
RPT Consulting Verified Certificate
Issued to all participants who meet the grading threshold. Results dashboard shared with ITER HR/L&D on request. Certificate is shareable on LinkedIn immediately post-programme.
60%+ — Passing Grade
70%+ — Certificate Issued
Before Day 1
Student Onboarding Checklist
Complete every item on this list before the first class. Click each item to mark it done. Arriving unprepared costs the entire cohort — not just you.
Python Environment
Python 3.10+ installed and working (
python --version)
pip and virtual environments working (
python -m venv .venv)
VS Code + Python extension installed
Can run a Jupyter notebook locally (
pip install notebook)
Accounts & Access
GitHub account created and profile set up
HuggingFace account created (huggingface.co)
Google account with Colab access (colab.research.google.com)
OpenAI API account (free tier is fine for M1)
Weights & Biases account (wandb.ai — free tier)
Pre-Study
Reviewed Python basics: functions, classes, list comprehensions, file I/O
Can explain what a vector is and what "cosine similarity" means
Read: "Attention Is All You Need" abstract (no need for the full paper)
Tried ChatGPT or Claude with at least one structured prompt using Role + Task + Constraints
Install Before Class 1
pip install openai anthropic langchain
pip install chromadb faiss-cpu sentence-transformers
pip install gradio streamlit fastapi uvicorn
Confirm:
import openai; openai.OpenAI() runs without error
Team & Admin
Join the programme Slack/WhatsApp group (link from instructor)
Submit Python baseline skill audit (link from RPT before Day 1)
Identify 2 potential capstone teammates (start thinking now)
Read the academic integrity / honour code policy
Day 1 Expectations
Laptop charged, charger brought — every class needs your machine running
Colab notebook open and ready at the start of every session
No passive watching — if you are not typing, something is wrong
You will be cold-called to demo your lab output — always finish the lab